JP2000196907A - Method for selecting color to be assigned to pixel and method for encoding and storing picture - Google Patents

Abstract

PROBLEM TO BE SOLVED: To precisely reproduce a color picture by designating a signal value corresponding to a maximum pixel cluster in order to select a pixel cluster including the largest number of pixels to assign to a block. SOLUTION: In a process for dividing a picture into plural blocks, each block is formed of plural pixels and each pixel has values expressing luminance, color difference and hue at a separated position in the picture. Then, in the process which identifies at least one pixel cluster in the block, the pixel cluster includes plural pixels having nearly the same values of luminance, color difference and hue. Then, a process designating a signal value corresponding to the maximum pixel cluster for selecting the pixel cluster including the largest number of pixels to assign to the block is provided. For example a pixel map 10 includes picture elements namely the pixels 14. The plural pixels 14 are divided into plural blocks 12.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

FIELD OF THE INVENTION The present invention relates generally to image processing and, more particularly, to color selection for pixels in a block used in block truncated coded image compression techniques.

[0002]

2. Description of the Related Art Data reduction is required in data processing processes, where there is an excessive amount of data for the actual application that uses the data. Digital images--i.e., Images that are discretized in both spatial coordinates and luminance levels as obtained by scanning--are often numerous, and thus are desirable objects for at least one form of data reduction. Become. To be precise, not only does the inconvenience given to the user be reduced by allowing data to be processed at a higher speed, but also more complex data can be processed without significantly increasing the image processing time.
For example, the number of bits required to accurately draw a detailed halftone image is many times that of a single black document on an unprinted page. Similarly, rendering a color image accurately requires much more data than its very detailed halftones.
Without any form of data reduction, processing of documents containing halftone and color images can take an unacceptably long time.

[0003] All pixels in each block must be represented by only one of the two colors assigned to the block. If the two colors selected to represent each block are not carefully selected, the colors in the final decoded output image will not accurately represent the colors of the original image. The present invention discloses a method and apparatus for accurately reproducing a color image by selecting a color to be assigned to a two-color block and then assigning colors to the pixels in the block.

[0004]

Known methods of selecting colors on a block-by-block basis have attempted to find two extreme colors in a single pass. The extreme colors obtained from this pass are the two colors assigned to the block. Each pixel in the block is assigned the color closest to the pixel's actual color. This method is acceptable for the most general case-that is, when the block has two colors and a clear boundary between them. Although the neutral colors appear exactly along their borders, it is not that these intermediate colors are used to find the color chosen to display the block, which means that the extreme colors give the most realistic look. Because it becomes. However, a problem arises when there are two or more primary clusters consisting of pixels of substantially the same color. Blocks in more than one cluster may appear in regions of the image that contain high frequency texture data or noise. In this case, the color chosen as the second color selected for the block will represent the noise rather than the color of interest in the region.
Assigning a noise color to the block results in more pixels sticking to that color in the encoded or decoded image, thereby amplifying the noise.

[0005]

SUMMARY OF THE INVENTION A method provided by the present invention for selecting a color to be assigned to a pixel of an image for subsequent decoding and printing by a marking device is: dividing the image into a plurality of blocks. Wherein each block is comprised of a plurality of pixels; identifying a plurality of pixel clusters in the block, wherein the pixel clusters include a plurality of pixels of substantially the same color; a maximum pixel including a maximum number of pixels Selecting a cluster and designating the color of the largest pixel cluster as a first assignment to a block;
Calculating the average color of the pixels outside the largest pixel cluster and designating the average color of the outside pixels as the next assignment to the block.

[0006] A method provided by another embodiment of the present invention for encoding an image at a first resolution and storing it in an image buffer at a second resolution comprises: dividing the image into a plurality of blocks; Identifying each of a plurality of pixel clusters in each of the plurality of blocks, each pixel cluster including a plurality of pixels of substantially the same color; Selecting the largest pixel cluster for each block containing the pixels of the maximum pixel cluster and assigning the color of the largest pixel cluster as a first assignment to the block; average color for the pixels in the block outside the largest pixel cluster Calculating the average color of the outer pixels as the next assignment for the block; and for subsequent decoding and printing by the marking device And a step of storing a plurality of blocks in the reduced capacity image buffer.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Referring to the drawings for describing the embodiments of the present invention, the present invention is not limited thereto. FIG. 1 shows a digital format for displaying an original image. Pixel map 1 of commonly used type
0 is shown. As shown, the pixel map 10 is:
It includes a plurality of image elements, or “pixels” 14. The plurality of pixels 14 are divided into a plurality of blocks 12 for processing in the present invention. In the following description of the invention, it is assumed that the pixel map 10 includes color data. However, it will be appreciated by those skilled in the art that the present invention may be adapted for use in systems that provide grayscale halftone data and is not limited to the reproduction of color documents. Similarly, the invention is applicable to obtaining grayscale or black-and-white documents and printing in color, and vice versa.
It is intended that all such alternative methods be encompassed and that the present invention not be limited to the description contained herein.

Referring to FIG. 2, the present invention will be schematically described. A plurality of pixels in a color image are often divided into “clusters”. A cluster of pixels is defined herein as a group of pixels represented by digital signals each having substantially the same color values (luminance, color difference, and hue). As shown, once block 12 is obtained in step 102, clusters in the block containing the maximum number of pixels (also referred to as maximum cluster, maximum number cluster, or maximum pixel cluster) are identified as shown in step 104. You.
Once the largest cluster is identified, locate the pixel 14 at or closest to the geometric center of the cluster, as shown in step 106, and use the digital value of that pixel as one of the block colors. select. Although embodiments of the present invention are described as choosing the value of the geometric center of the cluster, those skilled in the art will appreciate that a mathematical center, such as, for example, the mean or central signal value, could be chosen instead. Will do. In fact, if it were advantageous to do so, one could choose the minimum or maximum signal value of the pixels in the cluster. The signal values for all pixels in the cluster are set equal to the chosen value. Next, in step 108, the average color value of a plurality of pixels in the block not included in the largest cluster is calculated and selected as another color of the block. Thus, the signal values for all pixels outside the largest cluster are set equal to the average of these signal values. In this way, if there is only one other cluster, the average is equal to the average of that cluster, and the resulting coded image will be as if it were a single pass into the block. Is equal to the closest value. However, if there are two or more additional clusters, their values are averaged, thereby reducing the effects of noise signals that may be included in the image.

Referring to FIG. 4, in one embodiment for finding the maximum pixel count cluster, as shown in step 302, on a pixel in block 12 to locate the pixel value range in each cluster. Performing a histogram analysis is included. Note that with reference to FIG. 3, a 4x with pixels numbered as shown.
The four size blocks have pixel signals, and the pixel signals are 166, 1 (in the order of pixels 1 to 16), respectively.
58, 164, 14, 162, 167, 12, 8, 17
0, 204, 248, 251, 202, 209, 24
6, and 242 are assumed. Pixel 1,
2, 3, 5, 6, and 9 (166, 158, 1
64, 162, 167, and 170) are in the same range, and pixels 4, 7, and 8 having values of 14, 12, and 8, respectively, are in different ranges, 204, 20
Pixels 10, 13, having values of 2, and 209, respectively.
And 14 are in yet another range, 248, 251, 2
Pixels 11, 1 having values of 46 and 242, respectively.
2, 15, and 16 are in yet another, fourth range.
Histogram analysis has shown that the signal values of each of the four pixel value ranges, 0-89, 90-180, 181-230, 231-255, are acceptable in such cases. As an example, the color of a pixel is shown as a simple scalar value (ie, number). In practice, pixel values are usually three- or four-dimensional vector quantities. For this reason, the histogram has three or four dimensions.

Referring again to FIG. 4, upon completion of the histogram analysis, each pixel is assigned to a cluster corresponding to the signal value range (pixel value range) in which the pixel is located, as shown in step 304. Thus, the number of pixels in each range will be counted as shown in step 306. In the example shown here, 90
The second cluster with a value of ~ 180 is clearly the largest cluster.

Turning to FIG. 5, the details of another embodiment, including finding the largest cluster according to the present invention, will be described. As described above, the histogram analysis of the block is performed as shown in step 402. The counters x and y corresponding to the pixel value range and the pixel being processed are initialized as shown in step 404. Value v _y of the pixel p _y
Is obtained in step 406 and compared with the upper threshold k _x of the first range, as shown in step 408.
If v _y is less than k _x , the pixel is assigned to range x, as shown in steps 410 and 412, and the counter corresponding to that range is incremented. On the other hand, v _y
If is not less than k _x , the pixel value range counter x is incremented and v _y is compared at step 408 to the next range upper threshold. The pixel value v _y is compared to the upper threshold of each range until an appropriate range is found. When a pixel is assigned to a range, the counter associated with it is incremented.

The next step is to check whether the last pixel in the block has been processed, as shown in step 416. If not, the pixel counter y is incremented and the appropriate range of locations for the next pixel is found using the steps described above in steps 406-414. This continues until the last pixel of the block has been processed. In this state (step 416), the processing for the current block stops as shown in step 420. The counters associated with all ranges are reviewed to determine which ranges have the most pixels and identify the largest cluster (maximum number of clusters). If there are still other blocks in the image, those blocks can also be processed as described herein to find the largest cluster of pixels in that block.

The present invention has been described above using conventional histogram analysis for simplicity. In the preferred embodiment, a "fuzzy" histogram is used. In a conventional histogram, whenever a value is in the range corresponding to a bin, the value is assigned to the "bin",
The counter for the bin is incremented. The respective ranges for the bins do not overlap with each other, and the complete set of ranges constitutes a range of histogrammed values. In a fuzzy histogram as used in the present invention, values are assigned to all bins within a certain radius. In one dimension, this means that every bin within any range centered on the current color value will cause the respective counter to increment. Thus, if the range is +/− 10 and the bin width is 16,
Depending on the value of 33, ranges 16 to 31 and 32 to 47
Are incremented in bins 2 and 3, respectively. Also, depending on the value of 40, ranges 16 to 31, 32
Bins 2, 3, and 4 corresponding to 47 through 47, 48 through 63
Are respectively incremented.

In two dimensions, conceptually, the value is surrounded by a circle of predetermined radius, and any bin partially overlapped by the circle will cause its counter to increment. In practice, the values added to the histogram have finite precision, and each coordinate may be separated into a number of bins (representing the bin at the center of the circle) and an offset (representing a fractional offset within the center). is there. There is only a finite (generally decimal) number of unique offset occurrences. For example, using 17 bins, at coordinate x, the bin number is floor [(x + 8) / 17] (where floor (x) is an integer whose maximum is x) and the offset is x−17. Floor [(x + 8) / 17]. For an 8-bit integer, there are only 15 expected values at which an offset can be generated in this case. If the number of bins is a power of two, the bin number is indicated by the high order bit of "x + bin width / 2", while the offset is the low order bit of the equation. Since there is a small set of expected offsets, a list of neighboring values may be pre-computed for each offset. Thus, without calculating the distance at the time during the construction of the histogram, it is possible to increment the value of the central bin and all affected neighbors (values within the radius r of the input value). .

In three dimensions or larger, circles generalize to spheres or hyperspheres, making the use of table lookups to find a set of neighboring values more important.

In this manner, a fuzzy histogram is constructed by incrementing the counter for each bin, each containing the new value, as well as for all bins within any radius of the value. The main advantages of this method can be referred to FIGS. 6 (A) and 6 (B). Here, a set of two-dimensional values is drawn along with the histogram bin boundaries. FIG.
In (A) (prior art), the histogram count shows only the values in the individual bins. In FIG. 6B (the embodiment of the present invention), the count is calculated for a radius of one bin width. The largest cluster is identified on the lower right side of the upper left of each figure. The present invention aims to find the largest cluster quickly. Using the center of this maximum cluster as the value of one color, the average of a plurality of pixels of a color not present in the cluster is used as another color.

The speed at which the largest cluster can be found is an important feature of the present invention. In summary, in a preferred embodiment of the invention, the data structure is a three- or four-dimensional histogram indexed by the low order bits of the color. Each pixel area contains a list of all pixels to map, along with a count of the list size. To avoid quantization errors, each pixel is input into each range within any radius of the pixel's color. When a pixel is entered into the list, the count for the corresponding range is incremented, and if the maximum count is exceeded to some extent, the count and range are recorded. When all the pixels in the block have been input, the range with the largest count includes the largest cluster of pixels. The average of the colors for these pixels is used as the first color.

The remaining colors are averaged to form a second block color. A bitmap showing all the arrangements of the pixel positions is formed, and the pixels included in the average are deleted from the bitmap to make it easier to find the pixels belonging to the second cluster.

In another embodiment of the invention, the process is repeated to find the largest cluster of unclassified colors for blocks to which three or more colors are assigned. In this way, after reviewing the counters associated with all ranges, not only determine which range has the largest pixel to identify the largest cluster, but also one or more next largest clusters Determine similarly. The value of the pixel at the geometric center of these clusters, or the mean, center,
A value, such as a maximum or minimum mathematical value, can also be assigned to pixels within an individual cluster.

In yet another embodiment of the invention, when two colors are found, they are extrapolated along the line between them to provide a form of edge enhancement.
You can also change one color. Thus, if the two colors are a and b, the change value is (a + b) / 2−
t (ba) / 2 and (a + b) / 2 + t (ba) /
2 and certain values of t are slightly greater than 1.

Note that the above example is also used in a method of encoding an image at a first resolution and storing it in a reduced capacity image buffer at a second resolution (for subsequent decoding and printing by a marking device). . That is, the reduced capacity image buffer stores the plurality of blocks in which the maximum pixel cluster whose color is designated as described above is selected for each block.

[0022]

Since the present invention is configured as described above, a color image can be accurately reproduced by selecting a color to be assigned to a two-color block and then assigning a color to a pixel in the block. It has the effect of being able to.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a type of pixel map commonly used to display an image for use in the present invention.

FIG. 2 is a generalized flowchart showing the sequential steps for selecting a pixel color according to the present invention.

FIG. 3 is a detailed diagram illustrating one embodiment of some of the present invention, including finding the largest cluster of pixels in the present invention.

FIG. 4 is a detailed flowchart illustrating some alternative embodiments of the present invention including finding the largest cluster of pixels in the present invention.

FIG. 5 is a flowchart of a printing system to which the present invention can be applied.

FIGS. 6A and 6B are diagrams illustrating a method of executing a fuzzy histogram according to the embodiment;

[Explanation of symbols]

 10 pixel map 12 blocks 14 pixels

Claims (2)

[Claims] 1. A method for selecting a color to be assigned to a pixel in an image, comprising: a) dividing the image into a plurality of blocks, wherein each block is composed of a plurality of pixels, and each pixel is an image. B) identifying at least one pixel cluster in the block, wherein the pixel clusters have substantially the same luminance, color difference, and hue. C) selecting a pixel cluster that includes a maximum number of pixels and specifying a signal value associated with the maximum pixel cluster for assignment to the block. How to select colors to be assigned. 2. A method of encoding an image at a first resolution and storing it in an image buffer at a second resolution, comprising: a) dividing the image into a plurality of blocks, each block comprising a plurality of blocks. B) identifying a number of pixel clusters present in each of the plurality of blocks, wherein the pixel clusters include a plurality of pixels of substantially the same color; and c) a maximum number of pixels in the block. Selecting the largest pixel cluster for each included block and specifying the color of the largest pixel cluster for assignment to the block; and d) reducing the volume for subsequent decoding and printing by the marking device. Storing the plurality of blocks in an image buffer.